Your data. Buffered. Relayed.
Ready for AI.
JustinX connects MQTT brokers, Kafka topics, and webhooks to the AI coding tools your team already uses — so they can build live dashboards, alerts, and reports without you setting up the infra.
Start for freeThe pipeline
From data source to every screen. In milliseconds.
Stream Buffer
Cached. Ordered. Ready to serve.
Runtime Watchers
AI-generated TypeScript evaluates every message.
Slack
Email
Webhook
JiraStream Buffer
Cached. Ordered. Ready to serve.
Runtime Watchers
AI-generated code evaluates every message
SlackEmailWebhookOne data source can feed hundreds of clients simultaneously. No polling. No duplicated connections.
How it works
Three steps. Zero infrastructure.
Connect your data source
Paste a connection string. Pick your protocol. We handle the handshake.
Supported today: MQTT (v3.1.1/v5, TLS, auth), Kafka (SASL/SSL), Webhooks (POST with HMAC verification)
Connections are org-scoped. Each connection gets a dedicated consumer. TLS everywhere.
We buffer and relay
Messages hit our stream buffer within milliseconds. We cache them, order them, and hold them until your clients are ready.
Buffer: Redis Streams with configurable retention (default: 1 hour). Messages keyed by connection + topic.
No message transformation. We deliver what your broker sends, byte-for-byte. Schema is yours.
Your team builds on live data
Your AI tool — Cursor, Claude, Lovable — connects via MCP and can now read your live stream. Your team describes what they want in English. The AI builds it.
Delivery: WebSocket relay pushes messages to every connected client in real time.
Fan-out is server-side. One buffer, many readers. Clients don't need to know about your broker.
AI integration
Built for AI tools, natively.
MCP (Model Context Protocol) is an open standard that lets AI coding tools read external data sources. JustinX speaks MCP out of the box. That means when your team asks Claude or Cursor to “show me a dashboard of sensor data,” the AI can actually read the live stream — no custom API code required.
{
"mcpServers": {
"justinx": {
"url": "https://api.justinx.cloud/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}Watchers
Runtime monitors on your live data stream.
A watcher is a server-side TypeScript process that reads every message from your data stream and evaluates it — in real time. Your team describes what they want in English. The AI generates the code. JustinX deploys and manages it. No infrastructure. No engineering tickets.
Describe
“Alert me when efficiency drops below 88%”
AI generates code
TypeScript watcher via MCP
Deployed & running
Evaluates every message, fires actions
How a watcher works
Your AI tool calls create_watcher via MCP
Sends the generated TypeScript source code to JustinX
JustinX spawns an isolated runtime
Sandboxed V8 process with access only to your connection's stream
The watcher reads every message in real time
Blocking read on the stream buffer — zero polling, zero lag
When conditions match, it fires
Slack alert, email report, webhook call, Jira ticket — whatever the code says
Auto-restarts on crash. Logs captured. Config hot-reloadable. Managed by a reconciliation loop — no babysitting.
Your team says it. The AI deploys it.
“Alert me when any converter's efficiency drops below 88%”
Watches power-in vs power-out per converter module. Sends Slack alert when the ratio degrades. Caught PE-7 dropping from 92% to 85% over two weeks.
“Flag any EVSE where cable temp exceeds 65°C for more than 2 minutes”
Monitors dispenser cable temperature readings. Triggers alert with full thermal context — pump status, fan speed, ambient temp.
“Daily report: how many sessions hit target SOC vs stopped early”
Aggregates start/end state-of-charge across all chargers. Emails a PDF summary at 6am. Operations knows before the morning standup.
Each watcher runs server-side on JustinX. Your broker doesn't know. Your infra team doesn't know. It just works.
Use cases
Real data. Real operations. Real results.
EV Depot Charging
Power Module Monitoring
A managed services team connects their depot's MQTT broker — 200 chargers across 15 sites, each publishing converter efficiency, cable temperatures, insulation resistance, and cooling system state every 5 seconds. A watcher catches PE-7's efficiency degrading from 92% to 85% over two weeks. Another flags insulation resistance declining 9 kΩ/day — 11 days before the safety shutdown that would have grounded a bus.
Factory Floor
Vibration & Thermal Anomaly Detection
A plant operations team connects a Kafka topic carrying CNC machine telemetry — spindle vibration, bearing temperature, and motor current across 40 machines. A watcher flags Machine 12: vibration at 2.3x baseline for 45 seconds, bearing temp climbing 0.5°C/minute. The maintenance ticket is created before the operator notices.
Fleet Logistics
GPS + Cold Chain Compliance
A logistics company pushes GPS coordinates and refrigeration unit telemetry from 500 trucks via webhook. They ask Lovable to build a mobile-friendly map with live ETAs and cargo zone temperatures. A watcher fires when Truck 238's Zone B rises above 4°C for more than 10 minutes — the dispatcher reroutes before the pharmaceutical shipment is compromised.
Compare
How JustinX compares
| Capability | JustinX | Grafana | Retool | DIY |
|---|---|---|---|---|
| MQTT / Kafka ingestion | Built-in | Plugin required | Not supported | You build it |
| Real-time WebSocket relay | Built-in | Limited (polling) | Not supported | You build it |
| AI tool integration (MCP) | Native | None | None | None |
| Setup time | Minutes | Hours–days | Hours | Weeks |
| Non-engineers can build apps | Yes (via AI) | No (PromQL) | Partial | No |
| Server-side automations | Built-in | Alert rules only | Workflows | You build it |
| Multi-tenant / org-scoped | Yes | Manual | Yes | You build it |
Grafana and Retool are great tools — for different problems. JustinX is purpose-built for bridging streaming data to AI-powered app builders.
Architecture
What's under the hood
| Protocol ingestion | MQTT v3.1.1/v5, Kafka (SASL/SCRAM, SSL), Webhooks (HMAC-verified) |
| Stream processing | Redis Streams with configurable retention and consumer groups |
| Real-time delivery | WebSocket fan-out with per-client backpressure and reconnection |
| AI integration | Model Context Protocol (MCP) — open standard, tool-agnostic |
| Compute (Watchers) | Isolated V8 runtimes, server-side TypeScript, per-org sandboxing |
| Auth & tenancy | OAuth 2.0, scoped API keys, org-level isolation, RBAC |
| Storage | PostgreSQL for metadata, time-series retention for telemetry history |
| Encryption | TLS 1.3 in transit, AES-256 at rest, zero-plaintext logging |
Security
Safe enough for production from day 1.
End-to-end encryption. Scoped API keys. Audit logs. Your data never touches our application layer — we relay, we don't read.
SOC 2 Type II
Audited annually
ISO 27001
Certified ISMS
AES-256
At rest & in transit
GDPR
EU data residency
See it work with your data.
Connect an MQTT broker or Kafka topic. Watch messages flow. Free — no credit card.
Start for free